Contents lists available at ScienceDirect

Journal of Power Sources

journal homepage: www.elsevier.com/locate/jpowsour

Cold sintering and ionic conductivities of Na3.256Mg0.128Zr1.872Si2PO12 solidelectrolytes

Haoyang Leng, Jiajia Huang, Jiuyuan Nie, Jian Luo∗

Department of NanoEngineering, Program of Materials Science and Engineering, University of California, San Diego, 9500 Gilman Drive, La Jolla, CA 92093-0448, USA

H I G H L I G H T S

• The cold sintering process (CSP) densifies NASICON specimens at 140 °C.

• With identical heat treatments, CSP leads to>2-6X increases in conductivities.

• Post-CSP annealing at 800 °C markedly increases the grain boundary conductivity.

• Mg doping enhances the densification and conductivity of cold-sintered specimens.

• A new opportunity of low-T sintering of solid electrolytes and solid-state batteries.

A R T I C L E I N F O

Keywords:Cold sintering processNASICONIonic conductivitySolid-state batterySodium-ion battery

A B S T R A C T

The recent invention of cold sintering process (CSP) suggests a new opportunity to densify ceramic solid elec-trolytes at reduced temperatures. In the study, Na3.256Mg0.128Zr1.872Si2PO12 (Mg-doped NASICON) specimenswere cold-sintered at 140 °C to ∼83% of the theoretical density. In addition, post-CSP annealing, at tempera-tures lower than that is needed for the conventional sintering of NASICON, can substantially increase the ionicconductivity (with or without significant further densification). For example, the conductivity of cold-sinteredMg-doped NASICON reached> 0.5mS/cm after annealing at (as low as) 800 °C; in comparison, a dry-pressedspecimen exhibited virtually no densification at a higher temperature of 1000 °C with< 0.1mS/cm con-ductivity. Moreover, a high conductivity of ∼1.36 mS/cm has been achieved for a Mg-doped NASICON cold-sintered and subsequently annealed at 1100 °C (doubling the conductivity of a dry-pressed specimen sintered atthe identical condition). Further mechanistic studies showed that annealing at 800–1100 °C increased the grainboundary (GB) conductivities of cold-sintered specimens. The CSP opens a new window to sinter the “thermally-fragile” ceramic solid electrolytes (such as phosphates and sulfates). In addition, it not only provides energy andcost savings but also enables new fabrication routes to make solid-state batteries.

1. Introduction

Recently, the research of sodium-ion batteries has gained great at-tentions due to the abundant sodium resources and its lower cost incomparison to lithium [1–5]. Furthermore, the development of solid-state batteries has also attracted significant interests in both the aca-deme and the industry, where solid electrolytes, instead of liquidelectrolytes that are prone to the leakage, dendrite growth, and ex-plosion risks, are used to improve safety and (potentially) energydensity (e.g., via enabling the use of metal anodes) [6–8]. Specifically,solid-state sodium batteries can in principle have low costs and possesshigh energy density for large-scale energy storage applications, such asthose used in power grids [8]. Here, one key enabling technology is to

cost-effectively fabricate sodium-ion solid electrolytes with high con-ductivity and stability [8–11].

Specifically, solid-state sodium batteries with natrium (Na) super-ionic conductors (NASICON) were demonstrated to be promising en-ergy storage devices [8,12]. Recently, solid-state batteries with NA-SICON based composite polymer electrolytes have also been designedand fabricated, which exhibited high specific capacity and stable cycleperformance [13–15]. The NASICON materials with the general for-mula of Na1+xZr2SixP3-xO12 (0≤ x≤ 3) were firstly developed byHong and Goodenough ∼40 years ago [16,17]. The Na+ ionic con-ductivities of NASIOCN ceramic electrolytes are relative high comparedto other inorganic solid electrolytes at elevated temperatures [18,19].In addition, NASICON ceramic electrolytes possess a wide

https://doi.org/10.1016/j.jpowsour.2018.04.067Received 28 February 2018; Received in revised form 16 April 2018; Accepted 19 April 2018

∗ Corresponding author.E-mail address: jluo@alum.mit.edu (J. Luo).

Journal of Power Sources 391 (2018) 170–179

0378-7753/ © 2018 Elsevier B.V. All rights reserved.

T

electrochemical window, high Na ion transference numbers, and highthermal stability [14,15,19,20]. NASICON oxides are more air andmoisture stable than sulfides [12–15]. However, undoped NASICONstypically have moderate total, i.e., bulk plus grain boundary (GB), ionicconductivities at room temperature (∼0.1mS/cm) due to the poorly-conducting ZrO2 phases precipitated at GBs during the high-tempera-ture (typically ∼1200 °C) sintering [16–20]. Recently, aliovalentdoping has been reported to increase the ionic conductivities of NA-SICON by several different groups independently [21–25]. In one study,Samiee et al. reported that the 0.128 at. % Mg doped NASICON can

exhibit a high ionic conductivity around 2mS/cm at room temperature(with the optimal sintering process) [21]; furthermore, a careful com-bined experimental and modeling investigation contributed the en-hanced conductivity to the formation of a conductive secondary and/orinterfacial Na3PO4 phase at GBs, representing one example with a clearexplanation of underlying mechanisms of the so-called “doping” effects(that are in fact more of interfacial and microstructural, than the bulk,effects) [21].

However, the high processing temperatures (typically 1200 °C orhigher) and long durations of NASICON materials in conventional

Fig. 1. (a) Measured particle size distribution of the synthesized Mg-doped NASICON powder (where the solid curve represents a lognormal fitting). (b) Relativedensity vs. Mg atomic percentage curve for NASICON specimens cold-sintered at 140 °C for 1 h under a fixed pressure of 470MPa. (c) Relative density vs. appliedpressure curve for Mg-doped NASICON (Na3.256Mg0.128Zr1.872Si2PO12) specimens cold-sintered at a fixed temperature of 140 °C and a fixed holding duration of 1 h.(d) Relative density vs. sintering temperature curve for Na3.256Mg0.128Zr1.872Si2PO12 specimens cold-sintered at a fixed pressure of 780MPa and a fixed holdingduration of 1 h. (e) Relative density vs. annealing duration curves for Na3.256Mg0.128Zr1.872Si2PO12 specimens cold-sintered at a fixed temperature of 140 °C and afixed pressure of 780MPa. The estimated errors of measured densities are±1.5%. 30 wt % water was added in each cold-sintered specimen.

H. Leng et al. Journal of Power Sources 391 (2018) 170–179

171

sintering can cause the Na and P evaporation, which results in theformation of insulating ZrO2 secondary phase at GBs to lower the ionicconductivity [26,27]. The poor sinterability of NASICON and requiredprolonged sintering at ≥1200 °C also add significant costs and energyconsumption, as well as cause various technical and compatibility is-sues to make it more difficult to fabricate thin NASICON membranesand/or integrate them into solid-state battery technologies, particularlyfor mass production of grid- (or other large-) scale energy storage de-vices. Thus, there are great technological advantages to explore novelsintering technologies to reduce the sintering or heat treatment tem-peratures of NASICON materials, motivating this study.

Recently, an innovative ceramic processing technology, cold sin-tering process (CSP), was developed by researchers at PennsylvaniaState University, which densifies ceramics with the assistance of anaqueous solution at a temperature below 300 °C for a short duration(typically 15–60min) under an uniaxial pressing pressure(100–500MPa) [28]. In the CSP, the initial densification is likelycaused by particle rearrangements under the applied pressure with theassistance of the aqueous solution among particles, where dissolution ispromoted. Subsequently, a supersaturated solution forms among par-ticles and a dissolution-precipitation process occurs to provide masstransport (densification) at a moderate temperature under the appliedpressure, while the water gradually evaporates [28]. In many (most)cases, a post-CSP annealing is still needed to achieve the final propertyrequirements, but it can be conducted at a temperature lower that isrequired for conventional sintering of dry-pressed specimens, therebyproviding a technological advantage. In the last a couple of years, theCSP has been applied to densify a spectrum of functional ceramics andceramic-polymer composites, including ZnO [29], BaTiO3 [30], ZrO2

[31,32], KH2PO4 [33], V2O5 based composites [34], and LiFePO4 [35].Specially, Berbano et al. demonstrated that the CSP could densifyLi1.5Al0.5Ge1.5(PO4)3 (LAGP) solid electrolytes to around 80% of thetheoretical density at 120 °C for 20min, but post-CSP annealing isneeded to increase their conductivity [36].

In this work, we further apply the CSP to a Mg-doped NASICON(Na3.256Mg0.128Zr1.872Si2PO12) and successfully achieved ∼83% of thetheoretical density at as low as 140 °C for 1 h. Similar to LAGP [36],post-CSP annealing (at temperatures lower than the conventional sin-tering temperature) was found to be necessary and effective to increasethe ionic conductivity of cold-sintered NASICON; moreover, our further

mechanistic study showed that the increased conductivity was achievedby reducing the high GB resistance in the as-CSP specimens. Specifi-cally, this study has demonstrated that the CSP can lead to a>2-6Xincrease in the conductivity in comparison with the conventional (dry-pressed) specimen sintered at the identical (reduced) temperature foran identical duration. Thus, this study suggests a promising new op-portunity for applying the CSP to fabricate solid-state electrolytes andbatteries.

2. Experimental

2.1. Powder synthesis

The NASICON powders with the compositions ofNa3.256Mg0.128Zr1.872Si2PO12 (Mg-doped NASICON) and Na3Zr2Si2PO12

(undoped NASICON) were synthesized by a solid-state reaction methodusing stoichiometric amounts of Na2CO3 (Fisher Chemical, 99.5%),MgO (Alfa Aesar, 99.95%), ZrO2 (Fisher Chemical, Laboratory Grade),SiO2 (Alfa Aesar, 99.9%), and NH4H2PO4 (Sigma Aldrich, 99.99%).Noting that we also made several other NASICON specimens with dif-ferent Mg contents that are reported in Fig. 1(b), but all other Mg-dopedNASICON specimens reported and discussed elsewhere have the com-position Na3.256Mg0.128Zr1.872Si2PO12. The mixtures were milled inisopropyl alcohol using a planetary ball mill for 24 h. The mixed pre-cursors were dried in an oven at 85 °C for 12 h and subsequently cal-cinated in air at 1150 °C for 5 h. Finally, the calcined powders wereagain ball milled in isopropyl alcohol for 48 h to decrease the particlesizes.

2.2. Cold sintering process (CSP) and Post-CSP treatments

The CSP powders were prepared by mixing the (Mg-doped or un-doped) NASICON powder with 30 wt % deionized (DI) water in mortarand pestle for 2–3min. The prepared powder was transferred into a die(with a diameter of ¼ inch) and uniaxially pressed at 780MPa at roomtemperature (25 °C) for 2min for a pre-pressing step. Then, the die washeated to a preset CSP temperature (140 °C in most cases) with aheating tape (wrapped around the die) using a temperature controller.The specimen was kept at this CSP temperature (e.g., 140 °C) for a fixedduration (typically 1 h) at a fixed pressure (typically 780MPa). We also

Fig. 2. SEM micrographs of the fractured surfaces of Mg-doped NASICON (Na3.256Mg0.128Zr1.872Si2PO12) specimens (a, c) dry-pressed at 25 °C and (b, d) cold-sintered at 140 °C, respectively, both at 780MPa for 1 h.

H. Leng et al. Journal of Power Sources 391 (2018) 170–179

172

studied the effects of varying the CSP temperature (120–180 °C),duration (10–120min), and pressure (300–780MPa). All cold-sinteredspecimens were baked at 200 °C for 12 h to remove any residualmoisture.

Most cold-sintered specimens were subsequently annealed at600–1200 °C for 6–48 h in a box furnace with a heating ramp rate of5 °C/min. For comparison, conventional dry-pressed NASIOCN speci-mens were also prepared by using the identical pressure (780MPa) atroom temperature and subsequently sintered (a.k.a. annealed) atidentical temperatures for identical durations.

2.3. Characterization and measurements

The specimen densities were obtained by measuring the mass anddimensions of each specimen and the theoretical density of NASICON is∼3.27 g/cm3. The crystalline phases of samples were identified by X-ray diffraction (XRD) using a Rigaku diffractometer with Cu Kα radia-tion. The scanning electron microscopy (SEM, FEI Quanta 250) wasused to observe the microstructures of fractured surfaces.

The specimens for conductivity measurements were prepared bysputtering Pt blocking electrodes on both surfaces of the pellets. Thethicknesses of all the specimens were in the range of 1.0–1.3mm.

Electrochemical impedance spectroscopy (EIS) was carried out by usinga Solartron 1255B EIS analyzer in the frequency range from 106 to 1 Hzwith AC amplitudes of 25, 50, or 75mV to measure the conductivity atroom temperature (25 °C) or controlled low temperatures. The con-ductivity measurements were also carried out at 0 °C, −20 °C, −40 °C,−60 °C, respectively, in a cold chamber with a temperature controller(LR Environmental Equipment Company) in dry air. The GB, bulk, andtotal ionic conductivities were obtained by fitting Nyquist plots usingthe Z-View software (Scribner, Inc.). A DC polarization test (with anapplied voltage of 1V at room temperature) was conducted to measurethe sodium ion transference number by following the reported standardprocedure [19,20]. The conductivity was continuously measured untilthe current was stabilized. The tested Na3.256Mg0.128Zr1.872Si2PO12

specimen (CSP + 1100 °C × 48 h) was prepared by sputtering Ptelectrodes on both surfaces of the pellet (dimension: ∼1.2 mm thickand 6.0 mm in diameter).

Fig. 3. XRD patterns of the Mg-doped NASICON (Na3.256Mg0.128Zr1.872Si2PO12) specimens after CSP only (i.e., an as-cold-sintered specimen) and CSP plus subsequentannealing at six different temperatures, along with the reference Na3Zr2Si2PO12 pattern (PDF No. 01-070-0234).

H. Leng et al. Journal of Power Sources 391 (2018) 170–179

173

3. Results and discussion

3.1. Cold sintering: effects of the Mg content and CSP parameters ondensification

The particle size distribution of the synthesized Mg-doped NASICONpowders was obtained by SEM analysis and are shown in Fig. 1(a); themean particle size was measured to be 430 nm, with a standard de-viation of 160 nm. Fig. 1 further shows the effects of the Mg content andCSP pressure, temperature, and duration (holding time) on the densi-fication of NASICON materials. As shown in Fig. 1(b), Mg doping cansubstantially increase the densification of NASICON. On one hand, theCSP with 30wt % water at 140 °C for 60min was found ineffective todensify the undoped NASICON; it resulted in a relative density of∼68.8%, which is only slightly higher than the ∼66.2% relative den-sity achieved by dry pressing a specimen at the identical pressure of470MPa. On the other hand, 0.128 at. % Mg doping increased the re-lative density to> 75% at 470MPa, and 0.512 at. % Mg doping furtherincreased it to> 78%. To explain this observation, we should note thatthe enhanced densification in cold sintering must be enabled by thesolubility of the solid phase(s) in water [28]. It is known that the NA-SICON type electrolytes are relative stable in water at 60 °C, suggestinglimited solubilities [37,38]; however, it is yet unknown whether ahigher temperature and a higher pressure can promote more solubilityto enable cold sintering. A prior study showed that Mg doping couldlead to the formation of conductive Na3PO4 secondary and/or

interfacial phase [21] that has high solubilities in water, which maypromote cold sintering. Moreover, Mg doping can also promote atransformation of primary NASICON from the monoclinic to therhombohedral phase [21], which may also be a factor that changes thesolubility and cold sintering kinetics. Further studies are needed toinvestigate the exact underlying mechanisms.

Although a higher Mg content can further increase the densificationof NASICON in the CSP, 0.128 at. % Mg doped NASICON was chosen forour systematic CSP experiments because of its optimal conductivity (asa higher Mg content can lead to a phase transformation of the primaryNASICON phase from the high-conducting monoclinic to the low-con-ducting rhombohedral phase, according to a prior study [21]).

Fig. 1(c) shows the increasing densification with increasing pressurefor the CSP of Mg-doped NASICON (Na3.256Mg0.128Zr1.872Si2PO12) atthe sintering temperature 140 °C for 1 h. Specifically, the specimencold-sintered at 310MPa has a relative density ∼69.8% and a highrelative density of ∼82.8% was achieved at 780MPa. Thus, the appliedpressure of 780MPa was used in the subsequent CSP experiments.Furthermore, Fig. 1(d) shows the CSP temperature (in the range of120–180 °C) have little influence on the final density; the subsequentCSP experiments were conducted at 140 °C. Finally, Fig. 1(e) shows thedensity variation with the CSP duration; the relative density increasedfrom a duration (hold time) of 30–60min, but it did not increase furtherfor an even longer duration of 120min, presumably because the waterwas dried after ∼60min. We also note that baking of the 60-min cold-sintered specimen at 200 °C for 12 h resulted in no weight loss, in-dicating that it was dried. Hence, the subsequent CSP experiments wereconducted for a fixed duration of 60min (or 1 h).

In summary, a high relative density of ∼83% has been achieved for

Fig. 4. The (a) relative density and (b) total ionic conductivity (measured atroom temperature) of the cold-sintered and dry-pressed Mg-doped NASICON(Na3.256Mg0.128Zr1.872Si2PO12) specimens vs. annealing temperature curves. Allcold-sintered and dry-pressed specimens were annealed for 6 h.

Table 1Measured room-temperature total conductivity for undoped (Na3Zr2Si2PO12)and Mg-doped (Na3.256Mg0.128Zr1.872Si2PO12) NASICON specimens that wereeither cold-sintered or dry-pressed and subsequently annealed at differenttemperatures for different durations. All CSP specimens shown here were cold-sintered at 140 °C and 780MPa for 1 h.

Specimen Processing Condition σt (mS/cm)

RelativeDensity (%)

Mg-doped CSP only 0.041 82.9Mg-doped CSP + 600 °C × 6 h 0.028 81.8Mg-doped CSP + 700 °C × 6 h 0.12 82.3Mg-doped CSP + 800 °C × 6 h 0.529 81.8Mg-doped Dry-Pressed

(140 °C × 1 h) + 800 °C × 6 h0.075 72.4

Mg-doped Dry-Pressed (RT) + 800 °C × 6 h 0.061 70.3Mg-doped CSP + 900 °C × 6 h 0.606 80.4Undoped CSP + 900 °C × 6 h 0.267 77.3Mg-doped CSP + 1000 °C × 6 h 0.655 80.2Mg-doped Dry-Pressed (RT) + 1000 °C × 6 h 0.082 72.3Mg-doped CSP + 1100 °C × 6 h 0.737 83.3Mg-doped Dry-Pressed (RT) + 1100 °C × 6 h 0.218 75.4Mg-doped CSP + 1100 °C × 12 h 0.878 86.9Mg-doped Dry-Pressed (RT) + 1100 °C × 12 h 0.28 76.5Mg-doped CSP + 1100 °C × 24 h 1.013 88.9Mg-doped Dry-Pressed (RT) + 1100 °C × 24 h 0.388 80.2Mg-doped CSP + 1100 °C × 48 h 1.362 93.6Mg-doped Dry-Pressed (RT) + 1100 °C × 48 h 0.601 82.3Undoped CSP + 1100 °C × 48 h 0.506 91.6Mg-doped CSP + 1150 °C × 30 min 0.78 86.7Mg-doped Dry-Pressed (RT) + 1150 °C × 30 min 0.219 74.1Mg-doped CSP + 1150 °C × 1 h 1.045 89.2Mg-doped Dry-Pressed (RT) + 1150 °C × 1 h 0.261 79.9Mg-doped CSP + 1150 °C × 6 h 1.3 92.6Mg-doped Dry-Pressed (RT) + 1150 °C × 6 h 0.67 82.7Mg-doped CSP + 1150 °C × 12 h 1.45 94.3Mg-doped Dry-Pressed (RT) + 1150 °C × 12 h 0.876 87.1Mg-doped CSP + 1150 °C × 24 h 1.182 93.8Mg-doped Dry-Pressed (RT) + 1150 °C × 24 h 0.851 87.9Mg-doped CSP + 1200 °C × 6 h 1.406 98.1Mg-doped Dry-Pressed (RT) + 1200 °C × 6 h 1.31 98.6

H. Leng et al. Journal of Power Sources 391 (2018) 170–179

174

Fig. 5. Measured (a, c) relative density and (b, d) total room-temperature ionic conductivity of cold-sintered (CSP) and dry-pressed Mg-doped NASICON(Na3.256Mg0.128Zr1.872Si2PO12) specimens vs. the duration for annealing at (a, b) 1100 °C and (c, d) 1150 °C, respectively. The estimated errors of measured densitiesare± 1.5%.

Fig. 6. SEM micrographs of the fractured surfaces of (a) as-cold-sintered (CSP only) specimens and cold-sintered Mg-doped NASICON (Na3.256Mg0.128Zr1.872Si2PO12)specimens after subsequent annealing at (b) 800 °C for 6 h, (c) 900 °C for 6 h, and (d) 1100 °C for 48 h.

H. Leng et al. Journal of Power Sources 391 (2018) 170–179

175

Mg-doped NASICON (Na3.256Mg0.128Zr1.872Si2PO12) specimens cold-sintered at 140 °C for 1 h at 780MPa, which was selected as the stan-dard processing conditions for the subsequent experiments. The mi-crostructures of a specimen cold-sintered with this set of conditions

(140 °C, 1 h, 780MPa) and a benchmark specimen dry-pressed at780MPa for 1 h are shown in Fig. 2 for comparison. Sinteringmorphologies, including particle packing and the formation on sin-tering necks, are evident in the cold sintered specimen. Mg doping wasfound effective and essential for promoting densification in cold sin-tering of NASICON, which also increased the conductivity of the spe-cimen as discussed subsequently in §3.3.

3.2. The effects of Post-CSP annealing

Fig. 3 displays the XRD patterns of cold-sintered Mg-doped NA-SICON specimens before and after 6 h of annealing at 800, 900, 1000,1100, 1150, and 1200 °C, respectively. All (both as-CSP and“CSP + annealing”) specimens possess a primary crystalline NASICONphase similar to those reported previously [19–21]. Peaks from minorsecondary phases, such as Na3PO4 and ZrO2, are also labeled in Fig. 3.An unknown secondary phase was also identified in the specimensannealed at 1000 °C and 1100 °C. The presence of minor secondaryphases is commonly found in sintered NASICON specimens [19–21].

Fig. 7. (a) The total, GB, and bulk ionic conductivities (measured at room temperature) of cold-sintered Mg-doped NASICON (Na3.256Mg0.128Zr1.872Si2PO12) spe-cimens vs. annealing temperature curves. All specimens were annealed for 6 h. The measured total, GB, and bulk ionic conductivities of an as-cold-sintered specimen(CSP only) are also shown. (b) EIS plots (measured at 25 °C) of a CSP (only baked at 200 °C for 12 h) specimen and a CSP + 700 °C × 6 h specimen, along withcorresponding fitted results (represented by the solid lines) using the equivalent circuit shown in the inset. (c) Arrhenius plots of the measured total conductivities ofvarious Mg-doped NASICON specimens.

Table 2Measured room-temperature GB (σgb), bulk (σb), and total (σt) conductivitiesfor Mg-doped NASICON (Na3.256Mg0.128Zr1.872Si2PO12) specimens cold-sintered(at 140 °C and 780MPa for 1 h) and subsequently annealed at different tem-peratures.

Processing Condition σgb (mS/cm) σb (mS/cm) σt (mS/cm)

As-CSP 0.009 0.125 0.041CSP + 700 °C × 6 h 0.046 0.829 0.12CSP + 800 °C × 6 h 0.233 0.924 0.529CSP + 900 °C × 6 h 0.254 0.861 0.606CSP + 1000 °C × 6 h 0.281 0.911 0.655CSP + 1100 °C × 6 h 0.388 1.01 0.737CSP + 1150 °C × 6 h 0.596 1.533 1.3CSP + 1200 °C × 6 h 0.672 1.836 1.406

H. Leng et al. Journal of Power Sources 391 (2018) 170–179

176

The effects of post-CSP annealing on improving the density andionic conductivity of cold-sintered Mg-doped NASICON specimens wereinvestigated (Fig. 4; Table 1). Dry-pressed specimens were also an-nealed (sintered) at the same conditions and the data are plotted to-gether in Fig. 4 for comparison. Post-CSP annealing at 600–1000 °C hadno appreciable effect on densification. Moderate densification occurredat 1100 °C (for 6 h), while a relative density of ∼92.6% was achievedfor 6 h annealing at 1150 °C (Fig. 4(a)). After annealing/sintering at thenormal sintering temperature of 1200 °C for 6 h, ∼98% of the theore-tical density was obtained for both cold-sintered and dry-pressed spe-cimens; we thereby concluded that the CSP would not provide muchadditional benefits when the normal sintering temperature is applied(but CSP can indeed provide substantial benefits at reduced sintering/annealing temperatures, in comparison with dry-pressed specimens). Itis also interesting to note that the densities of cold-sintered NASICONspecimens decreased slightly at 900–1000 °C, which may be related tothe decomposition of the minor secondary phase(s) that may release

small amounts of various gaseous species [36].Fig. 4(b) shows the total ionic conductivities (measured at 25 °C) of

both cold-sintered and dry-pressed Mg-doped NASICON(Na3.256Mg0.128Zr1.872Si2PO12) specimens after annealing at differenttemperatures. Perhaps the most interesting result in Fig. 4 is that an-nealing at 800–1000 °C can substantially increase the conductivity de-spite of no significant increase in the specimen density. Specifically, theconductivity of cold-sintered Mg-doped NASICON reached>0.5mS/cm after annealing at as low as 800 °C for 6 h. Moreover, the total ionicconductivity of cold-sintered Mg-doped NASICON reached ∼0.66mS/cm at 1000 °C, which is eight times (8× ) higher than that of the a dry-pressed specimen sintered at the identical condition (∼0.08mS/cm).The conductivity reached ∼1.3mS/cm at 1150 °C, which doubles theconductivity of the dry-pressed counterpart sintered at the identicalcondition (∼0.67mS/cm). However, the benefit of CSP became insig-nificant at the conventional sintering temperature of 1200 °C.

For a further critical comparison, a specimen was prepared by dry-pressing Mg-doped NASICON at 140 °C for 1h (without water addition)and subsequent annealing at 800 °C for 6 h. The measured density ofthis specimen was ∼72.4% and the measured ion conductivity was∼0.075mS/cm, which are only slightly higher than those of the spe-cimen dry-pressed at room temperature and subsequently sintered atthe identical condition (i.e., ∼70.3% density and ∼0.061mS/cm, re-spectively); however, these values are still substantially lower thanthose of the cold-sintered specimen sintered at the identical condition(i.e., ∼81.8% density and ∼0.529mS/cm, respectively). This addi-tional comparison further and directly demonstrates that the wateraddition in CSP is essential to promote the densification of Mg-dopedNASICON specimen.

We further investigated the variations of the density and ionicconductivity of both cold-sintered and dry-pressed Mg-doped NASICONspecimens with the annealing duration at 1100 °C and 1150 °C, re-spectively. Both density and ionic conductivity increased with the in-creasing annealing duration (Fig. 5). Specifically, 30-min short an-nealing at 1150 °C can densify a cold-sintered specimen to ∼86.7% toachieve a total conductivity of ∼0.78mS/cm, being significantlyhigher than the 74.1% relative density and ∼0.22mS/cm total con-ductivity of a dry-pressed specimen sintered at identical condition(Fig. 5(c)). Moreover, a high density of ∼93% has been achieved for acold-sintered Mg-doped NASICON specimen annealed at 1100 °C for48 h (Fig. 5(a)) with high total ionic conductivity of ∼1.36mS/cm(Fig. 5(b)), which are substantially higher than those of a dry-pressedspecimen sintered at the identical condition (i.e., ∼82.3% density and∼0.60mS/cm, respectively). While the densification and conductivitycontinued to increase with increasing annealing duration, at least up to48 h, at 1100 °C, they both leveled off after ∼12 h annealing at 1150 °C.Representative SEM images of the fractured surfaces of an as-CSP spe-cimen as well as three cold-sintered specimens annealed at varioustemperatures are shown in Fig. 6.

To summarize the most significant observations, the total ionicconductivity of as-CSP specimen with ∼82.8% density is only∼0.04mS/cm, which can be increased by ∼3 times to ∼0.12mS/cmafter annealing at a temperature as low as 700 °C, or by > 10 times to∼0.53mS/cm after annealing at a low temperature of 800 °C. Theseannealing temperatures are substantially lower than the conventionalsintering temperature of NASICON, where no appreciable densificationcan occur. The underlying mechanisms were probed and discussedsubsequently. Moreover, higher conductivities and densities can beachieved at higher annealing temperatures (e.g., ∼0.8 mS/cm and∼87% density for a 30-min short annealing at 1150 °C or ∼1.36mS/cm and ∼93% density for long annealing at a lower temperature of1100 °C for 48 h), which are substantially higher than those can beachieved by conventional sintering of dry-pressed specimens at thesame conditions (e.g., ∼3.5× and ∼2×higher conductivities in theabove two examples, respectively). These discoveries suggest new op-portunities for fabricating solid-state batteries, e.g., by co-firing

Fig. 8. (a) EIS plots (measured at 25 °C) of cold-sintered undoped(Na3Zr2Si2PO12) and Mg-doped (Na3.256Mg0.128Zr1.872Si2PO12) NASICON spe-cimens, respectively, annealed at 900 °C for 6 h and 1100 °C for 48 h, respec-tively along with corresponding fitted results (represented by the solid lines)using the equivalent circuit shown in the inset. (b) The DC polarization plot of arepresentative Mg-doped (Na3.256Mg0.128Zr1.872Si2PO12) NASICON pellet(measured at 25 °C), indicating an ionic transfer number of> 0.997.

H. Leng et al. Journal of Power Sources 391 (2018) 170–179

177

electrolytes and electrodes at reduced temperatures and/or shorteneddurations in one step. However, there will be trade-offs among pro-cessing temperature/duration and achievable conductivity/density.Yet, for example, fabrication of NASICON materials with a> 0.5mS/cm conductivity without going beyond 800 °C is a new possibility un-attainable previously, which can enable new fabrication routes forsolid-state batteries.

3.3. The mechanisms of conductivity increases during low-temperatureannealing

The enhancement of ionic conductivity of cold-sintered Mg-dopedNASICON at 700–1000 °C cannot be explained by the density im-provement, since there is essentially little or no densification (sintering)at 1000 °C or lower temperatures (Fig. 4). XRD studies did not revealany significant changes in the primary NASICON phase (Fig. 3). Yet, theenhancement of total ionic conductivity started at 700 °C (by ∼3× )and became substantial at 800 °C (to> 0.5mS/cm, or a>10× in-crease from the as-CSP specimen). In contrast, the conductivity of a dry-pressed specimen sintered at 1000 °C for 6 h is only ∼0.08mS/cm, oronly ∼1/8th of the measured conductivity (∼0.66mS/cm) of a cold-sintered specimen annealed at the identical condition.

The heat treatment effects on ionic conductivity enhancement ofother solid electrolytes have been investigated and attributed to crys-tallization of amorphous insulating phases at GBs in several prior stu-dies [39–41]. The low ionic conductivity of cold-sintered NASICONspecimens (despite achieving ∼83% densities) could be related to thedissolution and precipitation of secondary phases (presumably with lowcrystallinity or being at least partially amorphous), which are known toform for NASICON in contact with water [42]. Small particles are evi-dent on the surfaces of the NASICON grains (i.e., at GBs) in a fracturedcold-sintered specimen in Fig. 6(a), which could be the secondary(amorphous or poorly-crystallized) phase precipitated around NA-SICON grains in the CSP; the numbers of similar small particles indeedreduced after annealing at 800 and 900 °C [Fig. 6(b–d)] and such smallparticles virtually disappeared after annealing at 1100 °C. Un-fortunately, the exact composition of these small particles observed inFig. 6(a) cannot be measured by energy-dispersive X-ray spectroscopyand NASICON is too beam sensitive for conducting high-resolutiontransmission electron microscopy (HRTEM). However, similar amor-phous secondary and/or interfacial phases have been found in variousother cold-sintered ceramics [30,32,34].

In addition, a prior study [21] showed that Mg doping enhanced theionic conductivity of NASICON by promoting the formation of moreconductive secondary Na3PO4 phase, which were also found in ourspecimens (Fig. 3). Moreover, Na3PO4 can transform from a tetragonalα phase to a cubic γ phase at ∼600 °C with a jump in its ionic con-ductivity [43]. The enhancement of ionic conductivity of γ-Na3PO4

with a cubic structure is attributed to the increased number of mobileNa+ vacancies [44]. Thus, this α-to-γ phase transformation in thesecondary Na3PO4 phase, in addition to possible crystallization of thesecondary amorphous (or poorly-crystallized) phase, may contribute tothe observed increased (mostly GB, instead of the bulk) conductivity ofour cold-sintered specimens during annealing at 700–1000 °C. Un-fortunately, the amount of the secondary phase is too small for ex-amining its structure or crystallinity directly with XRD, and NASICON istoo beam sensitive for HRTEM (as the intense beam in HRTEM wouldamorphize the specimen in seconds).

To test the hypothesized mechanisms, we separate the GB con-ductivities and bulk conductivities by fitting EIS plots with a brick layermodel [45]; the results are shown in Fig. 7(a) and Table 2. It is noticedthat the bulk conductivity is always higher than the GB conductivity inall specimens, which indicates the total ionic conductivity is mostlylimited by the high GB resistance. Specifically, annealing at 700 °Csubstantially increased both bulk and GB conductivities. Fig. 7(b) dis-plays the impedance spectra of the cold-sintered Mg-doped NASICON

specimens before and after annealing at 700 °C, where the two semi-circles of the EIS plots correspond to bulk and GB resistances [46]. Anequivalent circuit model (as shown in the inset in Fig. 7(b)) with two R-CPE components to represent the bulk and grain boundary contribu-tions, respectively, was used for data fitting.

Notably, the bulk conductivities of specimens annealed at700–1000 °C are essentially the same [Fig. 7(a)]. Annealing at 800 °Cfurther increased (only) the GB conductivity substantially in compar-ison with that of the specimen annealed at 700 °C (while there is es-sentially no difference in the bulk conductivities). Furthermore, an-nealing at higher temperatures 900 and 1000 °C, respectively, increasedthe total conductivity slightly by modestly increasing (only) the GBconductivity. Finally, both bulk and GB conductivities increased afterannealing at high temperatures of 1100–1200 °C. Arrhenius plots of themeasured conductivities of various Mg-doped NASICON specimens areshown in Fig. 7(c).

These observations are largely consistent with the above-discussedhypothesis that annealing at 800–1000 °C increased the total con-ductivity substantially [Fig. 4(b)] mostly through increasing the GBconductivity [Fig. 7(a)] via (presumably) crystallization of secondaryand interfacial amorphous (or poorly-crystallized) phase(s) and/or theα-to-γ phase transformation in the secondary Na3PO4 phase.

Finally, we also compared the measured conductivities of cold-sin-tered undoped vs. Mg-doped NASICON specimens (Table 1; Fig. 8(a),where the equivalent circuit model is shown in the inset in Fig. 8(a)). Itis noticed that Mg doping substantially increased the total ionic con-ductivity in all cases. For example, the measured total ionic con-ductivities of the cold-sintered undoped and Mg-doped NASICON spe-cimens, respectively, after 900 °C× 6 h annealing are ∼0.27mS/cmand ∼0.61mS/cm, respectively. Moreover, the measured total ionicconductivities of the cold-sintered undoped and Mg-doped NASICONspecimens, respectively, after 1100 °C×48 h annealing are ∼0.51mS/cm and ∼1.36mS/cm, respectively. In both cases, Mg doping increasedthe total conductivities by more than 2× . As demonstrated by a priorcoupled experimental and modeling study [21], Mg doping can increaseGB conductivities by forming a conductive secondary and/or interfacialNa3PO4 phase at the NASICON GBs, which are consistent with thecurrent observations using cold-sintered specimens (Figs. 3 and 8(a)).Fig. 8(b) further shows the DC polarization plot of a representativeNa3.256Mg0.128Zr1.872Si2PO12 pellet, where the sodium ion transfernumber was measured to be > 0.997, verifying that the conductivityof the specimen is mostly ionic.

4. Conclusions

Mg-doped NASICON (Na3.256Mg0.128Zr1.872Si2PO12) specimens werecold sintered at 140 °C to∼ 83% of the theoretical density. Subsequentlow-temperature annealing at 800 °C can help to substantially increasethe conductivity by mostly increasing the GB conductivity to achieveis > 0.5mS/cm. In contrast, a dry-pressed specimen exhibited vir-tually no densification at a higher sintering temperature of 1000 °C anda much lower conductivity of< 0.1mS/cm. Moreover, 30-min shortannealing at 1150 °C of the cold-sintered specimen can achieve ∼87%density and ∼0.8mS/cm conductivity, in comparison with ∼74%density and ∼0.2mS/cm conductivity of a dry-pressed specimen sin-tered at the identical condition. A high density of ∼93% and highertotal ionic conductivity of∼1.36mS/cm have been achieved for a cold-sintered Mg-doped NASICON specimen, in comparison with ∼82%density and ∼0.6mS/cm of a dry-pressed specimen, after annealing/sintering at the identical condition of 1100 °C× 48 h. These discoveriessuggest new opportunities/routes for fabricating solid-state batteries,e.g., by co-firing electrolytes and electrodes at reduced temperaturesand/or shortened durations, where trade-offs among processing tem-perature/duration and achievable conductivity/density should bemade. The successful fabrication of NASICON materials witha> 0.5mS/cm conductivity without going beyond 800 °C, for example,

H. Leng et al. Journal of Power Sources 391 (2018) 170–179

178

represents a new possibility and enabling capability that have not beenachieved before. In a broader context, the CSP technology provides anew opportunity to sinter the “thermally-fragile” ceramic solid elec-trolytes for both Li+ and Na+ (such as phosphates and sulfates withvolatile components and prone to the decomposition and/or pre-cipitation of harmful secondary phases at high temperatures), and it canalso provide significant energy and cost savings.

Acknowledgement

This work was partially supported by a Vannevar Bush FacultyFellowship sponsored by the Basic Research Office of the AssistantSecretary of Defense for Research and Engineering and funded by theOffice of Naval Research through grant no. N00014-16-1-2569. J. N. issupported by the Aerospace Materials for Extreme Environments pro-gram of the Air Force Office of Scientific Research (AFOSR) under thegrant no.FA9550-14-1-0174 and J.N. and J.L. also thank the AFOSRprogram manager, Dr. Ali Sayir, for his support and guidance.

References

[1] K.B. Hueso, M. Armand, T. Rojo, Energy Environ. Sci. 6 (2013) 734–749.[2] B.L. Ellis, L.F. Nazar, Curr Opin Solid St M 16 (2012) 168–177.[3] Z.L. Jian, W.Z. Han, X. Lu, H.X. Yang, Y.S. Hu, J. Zhou, Z.B. Zhou, J.Q. Li, W. Chen,

D.F. Chen, L.Q. Chen, Adv Energy Mater 3 (2013) 156–160.[4] P. Senguttuvan, G. Rousse, V. Seznec, J.M. Tarascon, M.R. Palacin, Chem. Mater. 23

(2011) 4109–4111.[5] Z.Z. Zhang, Q.H. Zhang, J.A. Shi, Y.S. Chu, X.Q. Yu, K.Q. Xu, M.Y. Ge, H.F. Yan,

W.J. Li, L. Gu, Y.S. Hu, H. Li, X.Q. Yang, L.Q. Chen, X.J. Huang, Adv Energy Mater 7(2017).

[6] E. Quartarone, P. Mustarelli, Chem. Soc. Rev. 40 (2011) 2525–2540.[7] F. Lalere, J.B. Leriche, M. Courty, S. Boulineau, V. Viallet, C. Masquelier, V. Seznec,

J. Power Sources 247 (2014) 975–980.[8] C. Zhao, L. Liu, X. Qi, Y. Lu, F. Wu, J. Zhao, Y. Yu, Y.S. Hu, L. Chen, Adv Energy

Mater 2018 (2018) 1703012.[9] W.D. Zhou, Y.T. Li, S. Xin, J.B. Goodenough, Acs Central Sci 3 (2017) 52–57.

[10] V. Palomares, P. Serras, I. Villaluenga, K.B. Hueso, J. Carretero-Gonzalez, T. Rojo,Energy Environ. Sci. 5 (2012) 5884–5901.

[11] D. Kundu, E. Talaie, V. Duffort, L.F. Nazar, Angew. Chem. Int. Ed. 54 (2015)3431–3448.

[12] K. Hayashi, K. Shima, F. Sugiyama, J. Electrochem. Soc. 160 (2013) A1467–A1472.[13] Z.Z. Zhang, K.Q. Xu, X.H. Rong, Y.S. Hu, H. Li, X.J. Huang, L.Q. Chen, J. Power

Sources 372 (2017) 270–275.[14] J.K. Kim, Y.J. Lim, H. Kim, G.B. Cho, Y. Kim, Energy Environ. Sci. 8 (2015)

3589–3596.

[15] Z.Z. Zhang, Q.Q. Zhang, C. Ren, F. Luo, Q. Ma, Y.S. Hu, Z.B. Zhou, H. Li, X.J. Huang,L.Q. Chen, J. Mater. Chem. 4 (2016) 15823–15828.

[16] H.Y.P. Hong, Mater. Res. Bull. 11 (1976) 173–182.[17] J.B. Goodenough, H.Y.P. Hong, J.A. Kafalas, Mater. Res. Bull. 11 (1976) 203–220.[18] J.W. Fergus, Solid State Ionics 227 (2012) 102–112.[19] A.G. Jolley, G. Cohn, G.T. Hitz, E.D. Wachsman, Ionics 21 (2015) 3031–3038.[20] S.F. Song, H.M. Duong, A.M. Korsunsky, N. Hu, L. Lu, Sci Rep-Uk 6 (2016).[21] M. Samiee, B. Radhakrishnan, Z. Rice, Z. Deng, Y.S. Meng, S.P. Ong, J. Luo, J.

Power Sources 347 (2017) 229–237.[22] S. Song, H.M. Duong, A.M. Korsunsky, N. Hu, L. Lu, Sci Rep-Uk 6 (2016) 32330.[23] Q. Ma, M. Guin, S. Naqash, C.-L. Tsai, F. Tietz, O. Guillon, Chem. Mater. 28 (2016)

4821–4828.[24] S. Chen, C. Wu, L. Shen, C. Zhu, Y. Huang, K. Xi, J. Maier, Y. Yu, Adv. Mater. 29

(2017) 1700431.[25] A.G. Jolley, G. Cohn, G.T. Hitz, E.D. Wachsman, Ionics 21 (2015) 3031–3038.[26] J.S. Lee, C.M. Chang, Y.I. Lee, J.H. Lee, S.H. Hong, J. Am. Ceram. Soc. 87 (2004)

305–307.[27] A. Ahmad, T.A. Wheat, A.K. Kuriakose, J.D. Canaday, A.G. Mcdonald, Solid State

Ionics 24 (1987) 89–97.[28] J.P. Maria, X.Y. Kang, R.D. Floyd, E.C. Dickey, H.Z. Guo, J. Guo, A. Baker,

S. Funihashi, C.A. Randall, J. Mater. Res. 32 (2017) 3205–3218.[29] S. Funahashi, J. Guo, H.Z. Guo, K. Wang, A.L. Baker, K. Shiratsuyu, C.A. Randall, J.

Am. Ceram. Soc. 100 (2017) 546–553.[30] H.Z. Guo, J. Guo, A. Baker, C.A. Randall, Acs Appl Mater Inter 8 (2016)

20909–20915.[31] H.Z. Guo, J. Guo, A. Baker, C.A. Randall, J. Am. Ceram. Soc. 100 (2017) 491–495.[32] H.Z. Guo, T.J.M. Bayer, J. Guo, A. Baker, C.A. Randall, J. Eur. Ceram. Soc. 37

(2017) 2303–2308.[33] H.Z. Guo, A. Baker, J. Guo, C.A. Randall, J. Am. Ceram. Soc. 99 (2016) 3489–3507.[34] J. Guo, H.Z. Guo, D.S.B. Heidary, S. Funahashi, C.A. Randall, J. Eur. Ceram. Soc. 37

(2017) 1529–1534.[35] J.H. Seo, J. Guo, H.Z. Guo, K. Verlinde, D.S.B. Heidary, R. Rajagopalan,

C.A. Randall, Ceram. Int. 43 (2017) 15370–15374.[36] S.S. Berbano, J. Guo, H.Z. Guo, M.T. Lanagan, C.A. Randall, J. Am. Ceram. Soc. 100

(2017) 2123–2135.[37] R.O. Fuentes, F. Figueiredo, F.M.B. Marques, J.I. Franco, Solid State Ionics 139

(2001) 309–314.[38] M. Guin, S. Indris, M. Kaus, H. Ehrenberg, F. Tietz, O. Guillon, Solid State Ionics 302

(2017) 102–106.[39] P. Johnson, N. Sammes, N. Imanishi, Y. Takeda, O. Yamamoto, Solid State Ionics

192 (2011) 326–329.[40] C.R. Mariappan, C. Yada, F. Rosciano, B. Roling, J. Power Sources 196 (2011)

6456–6464.[41] V. Thangadurai, W. Weppner, J. Solid State Chem. 179 (2006) 974–984.[42] F. Mauvy, E. Siebert, P. Fabry, Talanta 48 (1999) 293–303.[43] A. Ghule, N. Baskaran, R. Murugan, H. Chang, Solid State Ionics 161 (2003)

291–299.[44] J.T.S. Irvine, A.R. West, Mater. Res. Bull. 22 (1987) 1047–1054.[45] H. Chung, B. Kang, Solid State Ionics 263 (2014) 125–130.[46] X.X. Xu, Z.Y. Wen, Z.H. Gu, X.H. Xu, Z.X. Lin, Solid State Ionics 171 (2004)

207–213.

H. Leng et al. Journal of Power Sources 391 (2018) 170–179

179